“According to Anthropic, the primary danger of its Mythos model is not its ability to find any single vulnerability, but its capacity to find thousands of them at scale, overwhelming defensive capabilities.”